Lyceum

Lyceum runs open-source AI models through a serverless, OpenAI-compatible API billed per token, and offers dedicated endpoints, GPU virtual machines billed per second, and reserved GPU clusters. Fine-tuning of open models is listed as in early access.

Customers are teams that build with open models and need to run them on their own infrastructure terms; the website names business customers with a shared Slack channel and agreed response times. Revenue comes from per-token and per-GPU-hour usage, plus dedicated and reserved capacity priced per contract. Two customer stories are published (Albatross and SONALAB).

The company describes itself as a sovereign European AI inference provider and lists Berlin and Zurich as offices. Its legal entity is Lyceum Technology Germany GmbH, and the website gives 2025 as the founding year. Serverless models run in Paris and Finland, and the careers page lists 18 open roles in Berlin and Zurich.

Product: Serverless inference for open models via an OpenAI-compatible API, dedicated model endpoints, and GPU virtual machines and clusters